Beta calibration: a well-founded and easily implemented improvement on logistic calibration for binary classifiers

Meelis Kull, Telmo De Menezes E Silva Filho, Peter Flach

Research output: Chapter in Book/Report/Conference proceedingConference Contribution (Conference Proceeding)

239 Downloads (Pure)

Abstract

For optimal decision making under variable class distributions and misclassification costs a classifier needs to produce well-calibrated estimates of the posterior probability. Isotonic calibration is a powerful non-parametric method that is however prone to overfitting on smaller datasets; hence a parametric method based on the logistic curve is commonly used. While logistic calibration is designed for normally distributed per-class scores, we demonstrate experimentally that many classifiers including Naive Bayes and Adaboost suffer from a particular distortion where these score distributions are heavily skewed. In such cases logistic calibration can easily yield probability estimates that are worse than the original scores. Moreover, the logistic curve family does not include the identity function, and hence logistic calibration can easily uncalibrate a perfectly calibrated classifier.
In this paper we solve all these problems with a richer class of calibration maps based on the beta distribution. We derive the method from first principles and show that fitting it is as easy as fitting a logistic curve. Extensive experiments show that beta calibration is superior to logistic calibration for Naive Bayes and Adaboost.
Original languageEnglish
Title of host publicationProceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS 2017)
PublisherJournal of Machine Learning Research
Number of pages9
Publication statusPublished - 1 Apr 2017
Event20th International Conference on Artificial Intelligence and Statistics (AISTATS 2017) - Fort Lauderdale, Florida, United States
Duration: 20 Apr 201722 Apr 2017
Conference number: 20
http://www.aistats.org/index.html

Publication series

NameJMLR Workshop and Conference Proceedings
PublisherJournal of Machine Learning Research
Volume54
ISSN (Print)1938-7228

Conference

Conference20th International Conference on Artificial Intelligence and Statistics (AISTATS 2017)
Abbreviated titleAISTATS
CountryUnited States
CityFort Lauderdale, Florida
Period20/04/1722/04/17
Internet address

Structured keywords

  • Jean Golding

Fingerprint Dive into the research topics of 'Beta calibration: a well-founded and easily implemented improvement on logistic calibration for binary classifiers'. Together they form a unique fingerprint.

Cite this